Design an event analytics platform
Company: Notion
Role: Data Engineer
Category: System Design
Difficulty: medium
Interview Round: Onsite
Design a data platform for raw product events with this schema:
- `event_id`
- `event_type`
- `user_id`
- `user_setting` as a nested object
- `event_metadata` as a nested object
Build an end-to-end design that supports analytics and downstream reporting. Discuss:
- whether to use a data lake, a data warehouse, or both, and the trade-offs
- how events are ingested and whether Kafka is useful
- when to use streaming versus batch processing
- how jobs are scheduled and orchestrated
- how to design data quality checks
- how to model warehouse tables for events, users, and workplaces
- how to preserve historical changes in a user dimension table
- how to model the many-to-many relationship between users and workplaces
- the trade-offs of flattening nested object columns versus keeping semi-structured data
Quick Answer: This question evaluates a candidate's skills in data engineering and system design for event analytics, covering platform architecture, data ingestion and streaming versus batch processing, orchestration, data quality, and schema modeling for nested and historical data as well as many-to-many relationships.